Video: "NEW ChatGPT Work is the Claude Cowork Killer? (Full Breakdown)" by Julian Goldie on YouTube.
What actually happened on 9 July
The same-day timing was not a coincidence - both OpenAI and Anthropic had been working toward agentic tools that sit inside your existing workflows rather than sitting in a separate chat window. OpenAI shipped ChatGPT Work as a new mode inside the ChatGPT desktop app (which simultaneously merged with the Codex app). Anthropic, meanwhile, pushed Claude Cowork to mobile - it had been web-only, and the iOS and Android apps arrived the same morning.
The practical effect is that both companies are now competing for the same slot: the AI that actually does work across your files, calendar, email, and connected accounts, rather than just answering questions about them.
What ChatGPT Work actually is
ChatGPT Work runs on GPT-5.6 with Codex built into the same interface. You connect it to third-party services and it gathers context across them - pulling from your documents, project boards, and email threads before it starts a task. The design intent is that you give it a multi-step brief and leave it running, with the model checking its own progress and adapting when something does not work out.
In practice, it is strongest on the kind of tasks that have a clear deliverable and a defined set of inputs. Writing a structured report from a folder of meeting notes, drafting a proposal based on a set of product specs, processing a list of customer queries into categorised responses - that sort of thing. The Codex layer means it can also write and run code as part of the workflow, which gives it a broader scope than a pure writing tool.
That said, it is a first release. The third-party integrations are functional but not comprehensive, and anything that requires nuanced judgment - deciding which of two contradictory briefs to follow, recognising when the project has changed scope - still requires human oversight.
What Claude Cowork brings to the comparison
Claude Cowork has been the more established tool going into this comparison. It has a longer track record on complex writing and reasoning tasks, particularly the kind of open-ended work where the output needs to hold together logically across a long document. The mobile launch matters because it means Cowork is now genuinely accessible mid-day, not just at a desk.
Where Cowork tends to do better is on tasks that require consistent tone and structure - writing a series of related pages, maintaining a house style across a project, producing content that reads as if one person wrote all of it. ChatGPT Work is faster to set up, but Claude's output on prose-heavy work tends to need less editing.
Worth knowing: the underlying model matters less than you might think for most business use cases. The difference between GPT-5.6 and Claude Fable 5 on a well-specified brief is narrow. What Julian Goldie's breakdown makes clear is that the real differentiator is how each tool handles ambiguity - and on that front, Claude is still ahead.
What Julian Goldie's comparison actually found
Julian ran both tools through a set of identical tasks - content planning, structured output generation, and multi-step project work. His finding was that ChatGPT Work closes the gap significantly on task execution, particularly for structured workflows with clear inputs. It is faster to connect to existing tools and handles the Codex-powered coding tasks better than Cowork does natively.
For businesses that already live inside the OpenAI ecosystem - using GPT-5.6 for most things and Codex for automation - ChatGPT Work is a genuine step forward. But it does not yet match Claude Cowork's output quality on complex, multi-document writing tasks where the brief is open-ended rather than prescriptive. The honest read is that neither tool is clearly better for everything. They are better for different kinds of work, and which one to reach for depends on the job.
What this means in practice for a UK business owner
The short version: you do not need to switch tools. If you have already built a workflow around Claude Cowork, this release is not a reason to restart from scratch. ChatGPT Work is genuinely worth testing if you are already paying for a ChatGPT subscription and want to see what it can do in a more structured mode.
The more important point is that agentic tools - AI that does multi-step work autonomously across your systems - are now mainstream products, not beta experiments. Both of these tools are commercially available, both are priced for business use, and both will improve quickly over the next six months. If your business does not have a tested workflow for at least one of them by the end of the year, you will have fallen meaningfully behind competitors who do.
To be fair: most businesses are not yet using either. The gap between "I have heard of these tools" and "I have a working setup that saves me three hours a week" is still where most of the value is sitting unclaimed.
Where this connects to NordSys
If tracking this kind of AI-agent news makes you wonder whether one could actually run inside your business, that's exactly what our AI Agents do - named, briefed and managed for you, no setup fee, from £6 a day.
See our AI Agents →